nizima LIVE 2.7 Adds AI Assistant and Voice-Only Model Control
Live2D Inc. released version 2.7 of its VTuber tracking app nizima LIVE on August 12, 2026. The update adds an AI assistant feature that links a Live2D model to a chat AI service such as ChatGPT, Gemini, or a local LLM, allowing users to converse with the model by text or voice. The assistant's speech output uses the VOICEVOX series, with 31 characters and 82 total voices including Zundamon and Shikoku Metan. The model's mouth and body move in sync with the speech, and emotions estimated from the conversation are reflected in facial expressions. Responses can be output as subtitles to OBS for display on a streaming screen. The AI assistant function is currently available only in Japanese. The update also adds a voice tracking function that moves the model using only voice input, without a camera, for eye blinks, mouth movements, and full-body motion. Other additions include a time parameter function linked to real-world time, lip sync support for all models on macOS, mouse click assignment to shortcuts, and Google account login. New sample models for Zundamon and Shikoku Metan are included, illustrated by Sakamoto Ahiru.